Datawhale open source
Ng Enda ChatGPT course series, Chinese version
Background of the project
In early May, DeepLearning.ai founder Ng Enda and OpenAI jointly launched the classic course "ChatGPT Prompt Engineering for Developers" for introductory large-scale model learning, which quickly became a phenomenon-level course for large-scale model learning and gained extremely high popularity. Subsequently, Professor Ng Enda cooperated with LangChain, Huggingface and other institutions to jointly launch a number of in-depth learning courses to help learners comprehensively and deeply learn how to use large models and develop complete and powerful applications based on large models.
At the beginning of the launch of the "ChatGPT Prompt Engineering for Developers" course, the DataWhale team devoted itself to learning and open source construction, translated each course into a Chinese version, reproduced its code, and explored the Chinese Prompt on the basis of the course Realization and optimization. For each course, the effect comparison of Chinese and English bilingual Prompt has been realized, and the content optimization and update have been maintained at a relatively high frequency. So far, DataWhale has completed the Chinese version of six courses of the open source project "LLM Introductory Course for Developers" based on Wu Enda's large-scale model series courses, and Github has received 6K stars.
Open source address: https://github.com/datawhalechina/prompt-engineering-for-developers
Figure 1 Project homepage
PDF tutorial, online reading
In order to better help domestic learners learn large-scale model development and optimize the tutorial reading experience, the project team specially set up a tutorial compilation team, based on the original course content, from the perspective of text reading, to create PDF tutorials and online reading that are more suitable for domestic readers .
According to the reading characteristics of PDF tutorials and online reading, we have adjusted the organizational form of the tutorials, optimized the content presentation method, and optimized code specifications, text theory, picture display and other aspects to present an easy-to-understand version for readers as much as possible , Easy-to-read text tutorials. Starting from theoretical study, combined with code practice, we have enriched and optimized the content to a greater extent on the basis of source code, so that readers can learn all the content of the course by only reading text tutorials.
Figure 2 Screenshot of online reading
Online reading address: https://datawhalechina.github.io/prompt-engineering-for-developers/
PDF tutorial download address: https://github.com/datawhalechina/prompt-engineering-for-developers/releases/tag/v1%2C0%2C0
New course for August - Gradio and W&B
In early August, Mr. Wu Enda released two new large model courses: "Building Generative AI Applications with Gradio" and "Evaluating and Debugging Generative AI". One of the two courses starts from Gradio, a well-known framework for large-scale model development, and guides developers on how to use Gradio and build large-scale model applications based on Gradio; the other combines the popular deep learning visualization tool wandb to introduce methods and practices for evaluating and improving generative AI.
The project team also implemented the follow-up of the new course at the first time, produced a Chinese version of the tutorial, reproduced its code and provided Chinese examples to better help domestic developers learn.
"Building Generative AI Applications with Gradio"
Instruct developers how to use Gradio to quickly and efficiently build user interfaces for generative AI through Python interface programs. The main content includes: image summary and application generation, using a simple interface to complete NLP tasks, describing and generating games, and communicating with any LLM.
Figure 3 Gradio picture
Evaluating and Improving Generative AI
Combined with wandb, it provides a set of systematic methods and tools to help developers effectively track and debug generative AI models. The main content includes: wandb introduction, training and evaluating diffusion models, evaluating and tracking LLM, fine-tuning language models, etc.
Figure 4 W&B Directory
write at the end
This is a turbulent new era, and every minute and every second is full of new changes. AIGC is one of the most imaginative and creative futures. We want to bring the future to more people, so that everyone who is interested in this can learn and embrace the big model and the future it will create. We are not only translating, but also creating more thoughts and values from the Chinese community. Looking forward to the affirmation and feedback of every reader.
Due to the lack of time and limited energy of the creative team, some omissions and even mistakes in the tutorial are inevitable. We hope that learners can actively give us suggestions or directly contribute to the project while learning. Let us work together to polish the tutorial for future learners Provide better content.
Dry goods learning, like three times ↓